AI Can Help Differentiate Parkinson Disease From Essential Tremor

Investigators assessed different machine learning models to see which was best at distinguishing Parkinson disease from essential tremor.

Researchers in China developed and evaluated machine learning algorithms to better identify and classify Parkinsonian and essential tremors for diagnosis. The findings, published in Frontiers in Neuroscience, offer a reference for the intelligent diagnosis of Parkinson disease (PD) and “show promise for use in wearable tremor suppression devices,” authors wrote.

Because PD symptoms are complex and become more severe in later stages, early diagnosis of the condition and introduction of effective treatment are crucial to disease management.

However, overlapping features between PD and essential tremor (ET), such as tremors of the upper limbs, can make it difficult to distinguish between the conditions, and the task usually comes down to doctors’ clinical experience. According to authors, approximately one-quarter of patients with PD are misdiagnosed as having ET.

In addition, “Some efficient and accessible non-invasive biomarkers such as tremor signals including tremor acceleration and surface electromyogram (sEMG) have been investigated for the differentiation between PD and ET,” they wrote.

A total of 7 prediction models created to help clinicians distinguish between PD and ET were included in the analysis (random forest [RF], eXtreme gradient boosting [XGBoost], support vector machine [SVM], backpropagation neural network [BP], ridge classification [Ridge], logistic regression [LR], and convolution neural network [CNN]).

Researchers compared the relative importance of various upper limb postures, tremor features, and demographics in each disease’s diagnosis.

Nearly 400 patients were recruited between June and November 2020, each with confirmed ET or PD with upper limb tremors; tremor information was collected via a medical device system and assessed in 4 different postures for each patient.

While each patient was resting, stretching, winging, and vertically winging, the devices recorded acceleration and sEMG measurements, for a total of 40 tremor variables collected from each patient.

Eighty percent of data collected were used as a training set in each model, with the remaining 20% serving as a validation set; the proportion of patients with PD or ET was consistent in each set.

“The average value of AU-ROC [area under receiver operating curve], which was calculated 10 times, was used as an indicator to evaluate the model for determining the different parameter combinations for each model,” researchers added.

Of the 398 patients assessed, 257 had PD and 141 had ET.

Analyses revealed:

  • The ensemble learning models including RF and XGBoost showed the best overall predictive ability with accuracy above 0.84 and AUC above 0.90
  • The other 5 models lacked a significant predictive ability
  • The dominant frequency of sEMG of flexors, the average amplitude of sEMG of flexors, resting posture, and winging posture had a greater impact on the diagnosis of PD, whereas sex and age were less important

The relatively small sample size marks a limitation to the current analysis, while in future studies, more patients with ET will be necessary.

“With the further acquisition of data of ET subjects in future work, the performance of models will be further improved and more valuable results will be obtained,” authors concluded.

Reference

Xing X, Luo N, Li S, Zhou L, Song C, Liu J. Identification and classification of Parkinsonian and essential tremors for diagnosis using machine learning algorithms. Front Neurosci. Published online March 21, 2022. doi:10.3389/fnins.2022.701632